LEADERSHIP TURNOVER AND FOREIGN POLICY CHANGE: SOCIETAL INTERESTS, DOMESTIC INSTITUTIONS, AND VOTING IN THE UNITED NATIONS1

Royce Carroll Rice University rcarroll@rice.edu Brett Ashley Leeds Rice University leeds@rice.edu Michaela Mattes Vanderbilt University m.mattes@vanderbilt.edu October 5, 2012 ABSTRACT This study examines the effect of domestic political change on UN General Assembly voting. We argue that foreign policy change is most likely when a new leader comes to power who relies on different societal groups for support than her predecessor, but that the extent to which foreign policy behavior changes with domestic interests depends on domestic political institutions. Democratic political institutions should increase foreign policy consistency, even in areas that are less subject to executive constraints. We test our hypotheses using a new measure of UNGA voting patterns and new data on changes in leaders supporting coalitions. We find evidence in support of our hypotheses: change in the societal support base of leaders leads to change in UN voting, especially in non-democracies. This study lends credence to the perspective that foreign policy, like domestic policy, can vary with the particular interests that leaders represent.

Earlier versions of this paper were presented at the The International Politics of Autocracies workshop at Rice University, May 3-5, 2012 and the annual meeting of the American Political Science Association, Washington, DC, September 2-5, 2010. The authors names are listed in alphabetical order; they view their contributions as equal. The research was supported by NSF grant SES-0921830. The authors thank Josh Clinton, Courtenay Conrad, Cliff Morgan, Jim Morrow, Alexander Thompson, Erik Voeten, Jessica Weeks, and participants at the Rice workshop for comments on earlier versions of this research and Anna Carella, Daina Chiba, Matthew DiLorenzo, Jinhyeok Jang, Jesse Johnson, Naoko Matsumura, Mariana Rodriguez, and Eelco Van der Maat for research assistance. Upon publication, all data and analysis mentioned in the text and footnotes will be available at http://www.ruf.rice.edu/~leeds/.

I. INTRODUCTION What causes change in foreign policy? Certainly foreign policy is driven in part by the exigencies and opportunities presented to a state as a function of its power and geographic position and the structure of the international system. The opportunities and constraints of the international system are not always fully decisive, however, and thus foreign policy is also influenced by state preferences, which in turn are determined through domestic political processes. Change in foreign policy can result from changes in the international system or a states position within it, but change in foreign policy can also result from changes in domestic politics. Not only should changes in domestic political institutions affect foreign policy, but changes in which societal groups hold power may also affect the broad contours of a states foreign policy. The purpose of this study is to examine the effect of changes in domestic sources of leader support on foreign policy. Because leaders wish to stay in power, and because every leader relies on the support of some group of domestic actors to maintain her leadership position, leaders tend to pursue policies favored by their core societal support groups. While foreign policy may not be the most important factor dividing groups within society, we argue that often domestic cleavages do correspond to different foreign policy preferences. Thus, when a leader comes to power who depends on a different constellation of societal groups for support than her predecessor, policy change, including foreign policy change, becomes more likely. The level of change in foreign policy associated with changes in the domestic source of leader support is mediated, however, by domestic political institutions. Two factors in particular should affect the extent to which changes in domestic support result in significant changes in policy. First, the size and breadth of the support coalition necessary to keep a leader in power

affects the likelihood of foreign policy change. When leaders require the support of large proportions of society to stay in power, policy is more likely to focus on matters of general interest rather than particularistic policies, and thus should exhibit more stability (Bueno de Mesquita et al. 2003, McGillivray and Smith 2008). Second, the extent to which executives control policy independently affects the extent of possible change. To the extent that policymaking requires compromise with other political actors, policy stability is increased (Martin 2000, Tsebelis 2002). Since democratic leaders require the support of larger proportions of the population to stay in power and are more constrained by other actors, we expect democracies to exhibit more stability in foreign policy than non-democracies. To test our argument about broad changes in the foreign policy of a state, we make use of voting patterns in the United Nations General Assembly (UNGA). The UNGA considers a wide variety of issues, the combined effect of which should reveal many aspects of a states grand strategy, international alignments, and foreign policy proclivities. Interpreting this data using a spatial voting model that employs bridge votes to induce comparability across UNGA sessions, we measure changes in voting patterns during the 1985-2008 time period. Our novel measure allows us to capture over time changes in a countrys aggregate voting behavior and interpret positions on a consistent scale. Using these data, we compare voting patterns for state leaders to the voting patterns of their predecessors. We have developed rules for distinguishing leadership changes that maintain the same domestic source of leader support from those in which the leaders support base changes. Commensurate with our hypotheses, we find that changes in source of leader support are associated with changes in UNGA voting patterns, but that this effect manifests itself mainly in non-democracies rather than in democracies.

Predicting foreign policy change is important to policymakers, and it is also an area in which scholars have not been systematically successful. This study aims to shed light on the empirical relationship between domestic politics and foreign policy. Our results show not only that foreign policymaking is shaped by the same factors as domestic policymaking, but that the competition among domestic groups is a powerful predictor of foreign policy change, even compared to a factor traditional international relations scholars would emphasize the shift in the structure of the international system at the end of the Cold War. Changes in source of leader support are particularly likely to lead to a redirection of foreign policy in non-democracies. The foreign policies of democracies, on the other hand, are more resilient to changes in the source of leader support. Responsiveness and stability are not necessarily at odds in democracies. II. THE DOMESTIC BASIS OF FOREIGN POLICY For much of the history of international relations scholarship, analysts have implicitly or explicitly assumed that foreign policy is developed differently than domestic policy. The analysis of domestic policy focuses on the outcome of competition among groups acting through particular political institutions, while foreign policy is often assumed to be driven by international imperatives or a commonly accepted view of the national interest. Scholars recognize that domestic competition may influence low politics areas such as trade policy, but typically consider other areas of foreign policy to be outside the realm of political competition and not subject to partisan, sectoral, ethnic, or regional interests.2 In fact, theories of grand strategy and international alignments usually turn to domestic politics only as an ad hoc explanation for behavior that appears unusual from the standpoint of international variablesfor example, the isolationist turn of the United States between the two world wars. As Narizny

For a review of the literature linking domestic politics to trade policy, see Milner (1999). 3

(2007, 3) writes, The extant literature . views domestic politics as a constraint on the pursuit of the national interest, not as the fundamental determinant of state behavior. Recently, scholars have begun to challenge this view. Narizny (2003, 2007), for example, argues that class and sectoral interests influence alliance policy and grand strategy. According to Narizny (2003), owners of capital and labor experience different costs and benefits from military spending, leading owners of capital to prefer a multilateral internationalist strategy based on alliances and owners of labor to prefer a more isolationist strategy based on self help. Narizny (2007) also argues that economic sectors within a society that depend on international trade and investment prefer more internationalist policies. He goes on to claim that among the internationalist coalition those economic actors who interact mainly with other core states are likely to prefer law-based relations, whereas economic actors who interact mainly with the periphery are more likely to see force as a useful and appropriate means for arbitrating international issues. Importantly, these different economic interests may exist within a state simultaneously and compete with one another for control of policy (see also Lobell, 2004). The result is that changes in the coalitions that have power domestically can result in changes in foreign policy even in the area of high politics. Economic interests are not the only source of different foreign policy preferences within states. Davis and Moore (1997), for example, find evidence that when an ethnic minority has political influence in a state, that state is more likely to engage in international conflict with states repressing members of the same ethnic group. Regional, religious, and ideological divisions may also affect a states general foreign policy orientation. Consider the case of Ukraine, in which a major source of political competition is between ethnic Russians based primarily in the East and ethnic Ukrainians based largely in the West. While under Prime

Minister Yulia Tymoshenko Ukraine had actively pursued NATO membership, the election of Viktor Yanukovych in 2010 was followed nearly immediately by a halting of that process. In Iran, the replacement of the secular Muslim monarchy of Shah Mohammad Reza Pahlavi with an Islamic Republic under Ayatollah Khomeini led to a significant change in international alignments. While this transition undoubtedly included changes to the domestic institutional structure, the change in the interests and goals of the primary regime supporters was even more dramatic and played a crucial role in the re-direction of Iranian foreign policy. While certainly international relations scholars have embraced the importance of domestic influences on foreign policy, much of the systematic work on this subject has emphasized the influence of domestic political institutions. These institutions, which define the decision-making process and accountability structure for a regime, change little across time and best provide cross-sectional explanations for variation in foreign policy. By contrast, the policies pursued by leaders through these institutions reflect shorter-term change in the groups to whom they are accountable. Despite the prevalence of anecdotal and post hoc assessments that such factors have been crucially important in particular cases, the extent to which changes in societal groups with influence over policymaking lead to foreign policy restructuring is not well established outside the economic policy arena. In this paper, we build on the idea that despite the constraints and imperatives of the international system, leaders retain meaningful choices in the foreign policies they pursue. These choices are influenced by the interests and preferences of societal groups that the leader represents and the domestic institutional context. Below we elaborate our argument linking changes in the domestic groups providing support to a leader to changes in foreign policy.

III. DOMESTIC INTERESTS, INSTITUTIONS, AND FOREIGN POLICY CHANGE We assume that, within societies, there are more or less well defined groups of likeminded individuals who share political interests as a result of their economic, ethnic, religious, regional, or other characteristics and thus have similar policy preferences on at least some issues. These groups compete with one another for political influence. We further assume that leaders wish to stay in power and that maintaining power requires retaining the support of some subset of these societal groups. The number of supporters necessary to stay in power depends on the countrys political institutions. Yet, the number of necessary supporters is smaller than the total set of people with a say in choosing leaders. (In Bueno de Mesquita et al.s (2003) terms, the winning coalition is smaller than the selectorate.) Thus, a leader may be replaced by a successor who depends on the support of a different subset of the population for the retention of power even without significant institutional change. Given that different domestic groups may have divergent preferences regarding international issues, we expect that foreign policy changes are most likely when a leader who depends on the support of a different societal group, and thus caters to different interests and preferences than her predecessor, comes to power. On the other hand, when a leader relies on the same source of support as her predecessor, it should be in her interest to continue to pursue similar policies. Compare, for example, the influence of the transition from Fidel Castro to Ral Castro on Cubas foreign policy to the transition from Fulgencio Batista to Fidel Castro. While we have seen little foreign policy change in Cuba since Ral Castro took power in 2008, the transition from Batista to Castro resulted in significant change in Cubas foreign policy. Batista supporters included business elites, land owners, the military, professional associations, and organized industrial laborers in Havana (Dominguez 1978). Among the

business elites, particularly influential were sugar mill owners and growers many of them Americans who favored positive relations with the U.S., the main market for their product. Batistas policies catered to these groups, and he pursued a staunchly pro-American and anticommunist diplomacy. Castros support base, on the other hand, included those who had previously been marginalized: farm workers, the unemployed, and peasants (Dominguez 1978). These groups demanded a re-structuring of the economy and the state and there were pressures from below that, though poorly articulated, nonetheless had a strategic impact upon Cuban relations with the United States (Dominguez 1978, 142). Castro established diplomatic relations with the USSR and China in early 1960, and in September Castro delivered a speech at the UNGA denouncing American imperialism (Moore and Pubantz 2008, 97). Whether a change in the domestic source of leader support results in noticeable change in foreign policy, however, depends on the domestic institutional context. There are two important sets of political institutions within states that influence the extent to which new leaders representing different core societal interests are likely to redirect foreign policy: those affecting the rules of leader selection and those affecting the rules of policymaking. Bueno de Mesquita et al. (2003) argue that rules of leader selection that require leaders to gain the support of large numbers of citizens (common in democracies) encourage leaders to pursue policies that provide public goods. Leaders operating in systems in which only a small number of supporters is necessary to stay in power, however, benefit most from providing private goods to their winning coalition. The incentive to produce broader public goods creates consistency in the types of policies pursued by different democratic leaders. On the other hand, when a new leader in a small winning coalition state (generally a non-democracy) comes to power, there may be very little overlap in the particularistic policies preferred by her small coalition compared to her

predecessors. Compared to democracies, we expect to see greater change on average between the policies of leaders with different support coalitions when those leaders depend on the support of a few to stay in office (McGillivray and Smith 2008). While the rules of leadership selection influence the extent to which leaders are likely to pursue broad-based public policies as opposed to the provision of narrowly-based private goods, rules governing the policymaking process have an influence on the ease with which a new leader can implement the policies she prefers. Some chief executives are relatively unconstrained in their ability to make and change policy. On the other hand, many chief executives operate in systems in which the consent of other political actors is necessary for the formulation of policy. Legislatures often have influence over the appropriation of funds, personnel appointments, the implementation of executive directives, and treaty ratification (Martin 2000). Similarly, coalition partners and subnational leaders often have a say in policymaking and a role in mobilizing resources for policy implementation. When leaders have to compromise with a number of other political actors in order to change policy, we are likely to see a bias towards the status quo even given a change in leader preferences. Democracies generally combine features of leadership selection rules and policymaking rules that discourage dramatic change in policy. Thus, given a change in the core societal supporting coalition of a leader, we expect more subsequent change in foreign policy in nondemocratic systems than in democratic systems, both because leader preferences are likely to be more particularistic and because leaders are less likely to be constrained from changing course by other political actors. That being said, in all states, we believe that change in foreign policy is more likely when new domestic groups gain influence over policymaking than it is when leaders depend on similar domestic groups to retain power.

IV. DOMESTIC POLITICAL CHANGE AND VOTING IN THE UN GENERAL ASSEMBLY An arena in which we may be able to witness and measure changes in the general foreign policy orientation of states is voting in the United Nations General Assembly (UNGA). The UNGA is the only forum in which all countries have the opportunity to express their views on issues that confront the international community. While wealthy and powerful countries are able to communicate their foreign policy positions in a number of ways (through foreign aid, economic sanctions, or the use of force, for instance), the foreign policy tools of developing states tend to be more limited. For them, the UNGA annual meetings provide a particularly valuable opportunity for making their views heard (Vengroff 1976, Moon 1985). The UNGA covers a wide range of international issues such as the North-South gap, disarmament, and human rights. Occasionally topics covered in the UNGA are directly relevant to domestic constituencies and different societal groups take different stances on these issues (Moon 1985).3 Many issues considered by the UNGA, however, do not map directly to domestic cleavages. On these issues, large portions of the domestic population are likely indifferent to the specific topics under consideration in the UNGA. Thus, we do not see UNGA voting as directly reflecting the interests of domestic groups with regard to particular votes. Instead, we view UNGA voting patterns as providing an indirect reflection of the foreign policy position of a state that corresponds to the broad preferences of domestic groups for alignment with or distance from particular states. Many UNGA resolutions, for example, represent
3

While Moons discussion focuses on developing countries, there are notable differences on

some of these policy issues in developed democracies as well. Consider for example the differences among West German voters, and their parties, on questions of nuclear proliferation during the 1980s. 9

declarations of affinity or disagreement with other states and governments votes on these topics will reflect their constituencies preferences of who to ally with and who to oppose on the international stage. In the aggregate, a countrys voting record in the UNGA is a latent indicator of its grand strategy and international alignments; it is a record of how the state wants to be seen by others and the positions it is willing to take publicly on a wide variety of issues. UNGA voting is particularly informative regarding domestically motivated changes in a countrys foreign policy positions for three reasons. First, unlike treaty cooperation, UN voting is not regulated by international law and countries are free to revise their positions as frequently and as much as they desire. Second, the non-binding nature of most UNGA resolutions makes it more likely that a countrys overall behavior can reflect a reaction to the views of important domestic constituencies rather than international strategic considerations (Gartzke 1998, Voeten 2001). While some scholars have argued that UNGA votes may be bought, the conditions under which aid or loans influence UNGA voting appear to be limited, and evidence concerning the conditions under which this phenomenon occurs is mixed.4 Moreover, the willingness to sell ones vote is also an indication of foreign policy, and thus even if such transfers occur, they
4

For example, Lai and Morey (2006) claim that non-democratic aid recipients increase their

voting similarity with the U.S., while Carter and Stone (2011) argue that it is democracies that vote with the U.S. in return for economic support. Studies show that countries are only more likely to vote in line with the U.S. if they receive World Bank non-concessional loans (Dreher and Sturm 2012) or benefit from general budget aid and untied grants (Dreher, Nunnenkamp, and Thiele 2008), and not in return for other kinds of loans and grants. Wang (1999) argues that aid affects UNGA voting only on the small proportion of votes that the U.S. considers to be particularly important. 10

would not undermine our claim that the aggregation of UNGA votes serves as a representation to the world of a states chosen policy positions.5 Third, unlike many other foreign policy behaviors, UN voting does not strictly involve interdependence among states. For trade to occur, an alliance to form, or a war to start, at least two states must choose to participate. UNGA voting, on the other hand, may involve consultations with other countries but ultimately reflects autonomous decisions by each government. Thus UN voting patterns can more easily change as the result of domestic political shifts in a state. The universal membership of the UNGA, the opportunity it provides to all states to express their positions on different international issues, the lack of international legal constraints, and the fact that voting is a monadic activity make the UNGA an excellent forum for studying countries foreign policy positions. In fact, the extent to which two countries vote together on UNGA resolutions forms the basis for a commonly used measure of interest similarity and international alignment (Gartkze 1998). We are not the first to consider the influence of domestic political changes on UN voting.6 Early studies by Vengroff (1976) and Moon (1985) focus on the effect of domestic
5

A similar logic applies to other types of insincere votes, such as the decision of authoritarian

governments to vote in favor of human rights resolutions. While such a vote does not mean that the government and its domestic elites are truly committed to improve human rights conditions, it does reflect an underlying preference to be seen as cooperating with Western values and this is an important component of a countrys overall foreign policy that can be affected by preferences of domestic elites to maintain business ties to the West.
6

There are also numerous studies that more generally examine the issue dimensions that

dominate voting in the UNGA, identify voting blocs and their membership, and determine 11

instability, especially violent leader transitions, on UN voting and report mixed findings. Hagan (1989) expands the scope of domestic changes to include non-violent changes in leadership that bring to power different groups from the same or opposite end of the political spectrum and shows that both revolutionary and non-revolutionary changes influence Third World countries alignment with the U.S. between 1946 and 1984. Voeten (2001) moves beyond Hagans study by expanding the empirical domain to all countries, but at the same time returns to a more limited focus on how institutional shifts affect foreign policy. Finding little support for his hypotheses, he concludes that his results do not necessarily demonstrate the irrelevance of domestic politics.7 Rather, he points out that the current dominant focus in the international relations literature on the effects of domestic institutions is too restricted. The interaction between institutional arrangements and societal preferences produces political outcomes (Voeten 2001: 31). Recent studies have picked up on the notion that societal preferences and the resulting policy orientation of governments should affect a countrys UN voting record. Potrafke (2009), for example, shows that OECD countries are less likely to vote similarly to the U.S. when their leaders are from left-leaning parties, and that this relationship is particularly strong when the U.S. President is a Republican. Dreher and Jensen (2012), in turn, demonstrate that when a states leader is of similar political orientation (that is, left or right) as the U.S. President, the country is more likely to vote similarly to the U.S. whether voting dimensions and bloc membership have changed over time (e.g. Alker and Russett 1965, Kim and Russett 1996, Voeten 2000, Voeten 2004).
7

A recent study by Ratner (2009) finds that democratic regime transitions lead to greater foreign

policy alignment with the U.S., but only if the U.S. had not previously supported the countrys non-democratic leaders. 12

While Hagan (1989), Potrafke (2009), and Dreher and Jensen (2012) direct attention to the importance of domestic preferences and government policy orientation, they do not consider the constraining effect of different domestic political structures. Yet, like Voeten, we believe that state policy derives from the interaction between domestic interests and institutions. Our study thus sets out to connect these two components of the domestic political process and to examine their effect on voting in the UNGA, using a larger sample than previous studies as well as a new measure of changes in domestic preferences and of UNGA voting. Unlike previously used measures of UNGA voting, our measure does not depend on comparison between the voting records of two states and thus allows us to disentangle which side is changing policy when two states move apart. In addition, our measure is designed to be less sensitive to apparent changes voting behavior that might only reflect differences in issues being considered by the UNGA from one session to the next. One thing to keep in mind when assessing the interaction of domestic political preferences with institutions is that, in UNGA voting, the influence of executive constraints should be somewhat muted in comparison to other policies like treaty formation or abrogation, the waging of war, or trade policy.8 Most UN votes do not require much consultation with other domestic actors directly. On the other hand, the appointment of UN ambassadors may be subject to legislative confirmation or compromise with coalition partners, and many UN votes do pave the way for foreign policy actions that require legislative action, for example treaty ratification or
8

Leeds, Mattes, and Vogel (2009) find that the effect of changes in the domestic sources of

leader support on the decisions of states to terminate alliances in violation of their terms is conditional on political institutions. Democracies show little foreign policy change with changes in the societal base of leader support in a highly institutionalized arena of foreign policy. 13

budgetary appropriations. Furthermore, leaders who have difficulty engaging in political action without the support of other actors will avoid casting unpopular votes in the General Assembly in order to ensure domestic veto players support on issues where they do have a gate-keeping role. Thus, the influence of executive constraints on UNGA voting should be lessened in comparison to some other foreign policy actions, but not absent entirely. Since democratic leaders both operate under conditions of higher executive constraint and derive their policy preferences from the need to satisfy larger constituencies, we believe democracies should exhibit more stability than non-democracies in UNGA voting patterns, but the distinction should be less stark than in policy arenas in which other political actors have more direct influence on the policy process, for instance, treaty formation. In the next sections we describe our research design and test the following hypotheses: H1: Leadership changes accompanied by changes in domestic sources of leader support are associated with more change in the states UNGA voting pattern than leadership changes that are not accompanied by changes in domestic sources of leader support. H2: Democratic states show more consistency in UNGA voting patterns across all leadership changes than non-democratic states. H3: The impact of changes in domestic sources of leader support on changes in UNGA voting patterns is stronger in non-democracies than in democracies. V. RESEARCH DESIGN Our goal is to examine the impact of changes in leadership on foreign policy, and specifically on UN voting. Following a number of recent studies, we adopt the leader as unit of analysis in investigating foreign policy behavior (Bueno de Mesquita et al. 2003, McGillivray

14

and Smith 2008, Goemans, Gleditsch, and Chiozza 2009). Thus, we compare the UN voting pattern of each leader during her entire tenure to her predecessors record during his time in office. We believe there are a number of advantages to this approach. First, it moderates the effects of any short term spikes during a leaders tenure that might be driven by exogenous factors that we do not control for. Second, a leader-based approach allows us to avoid imposing a time range for how long it takes for a new leaders foreign policy to become apparent; while in some systems change is immediate, in other systems it takes a while for the new leader to place new representatives in policymaking positions and implement her preferred policies. Third, it allows for a straightforward comparison of leadership transitions that are accompanied by changes in the societal groups that support the leader and leadership transitions without such changes. This provides evidence as to whether changes in supporting coalitions affect foreign policy beyond the effect that leadership transitions alone might have. Using information on leaders and the dates they were in office from Archigos (Goemans, Gledistch, and Chiozza 2009), we compare the mean of a leaders ideal point across UN sessions during her tenure to that of her predecessor.9 Since voting in the UNGA typically occurs late in the year (October through December)10, the countrys voting record in leader transition years is

Archigos covers all independent countries from 1875-2004. Leadership transitions from 2005-

2008 are coded based on worldstatesmen.org. If a leader briefly loses power during the year but then resumes office in time to affect UN voting we code the leader spell as continuing throughout the year. This ensures that consecutive leader spells represent different leaders.
10

Based on the Voeten and Merdzanovic (2008) data, we find that, during the 1985-2008 period,

December is the month with the highest number of votes (84%), followed by November with 15

attributed to the new leader rather than her predecessor. The exceptions are leader transitions that occur in the month of December. If a new leader comes to power after November 30th, it is unlikely that she can assemble a team of UN diplomats and provide comprehensive instructions in time to shape voting in the UNGA for that year. Thus, in the case of December leader transitions we attribute the countrys voting record in that year to the leader in power before December 1st. If there are multiple leader transitions in a given year, we code the leader in power during November as the leader in charge of the countrys voting in that year. Our data include a total 773 leaders over the 1985-2008 period. However, because we are interested in the change in the leaders ideal point from her predecessors, the first leader for each state during our observation period is dropped from the analysis.11 We also drop leaders from countries that did not vote in the UNGA.12 Finally, we exclude interim leaders who do not represent any particular societal groups but rather are tasked with maintaining the status quo until a new regular leader takes office. This leaves us with 541 leaders from 147 countries.13

12% of votes, and then October with about 2%. The remaining 1.8% of votes occur outside the regular session in the months January-September.
11

15 countries were dropped entirely because they only had one leader throughout the 1985-

2008 period. These countries are Kosovo, Equatorial Guinea, Guinea, Cameroon, Gabon, Somalia, Eritrea, Angola, Zimbabwe, Libya, Egypt, Yemen, Oman, Uzbekistan, and Kazakhstan.
12 13

This includes four states: Montenegro, Kosovo, Taiwan, and East Timor. 21% of countries have two regular (i.e. non-interim) leaders between 1985-2008, 17% have

three different leaders, 17% have four leaders, 19% have five leaders, and about 25% have six or more leaders. The average number of years a leader controls voting in the UNGA is 4.36. 16

Measuring Foreign Policy Changes Using UNGA Voting Our interest is in detecting changes in foreign policy that may arise from changes in political leadership. While individual UNGA votes do not necessarily capture specific foreign policy action, we argue that the set of votes cast by a leader in the aggregate captures latent tendencies in a nations grand strategy and international alignments, the notion of foreign policy emphasized here. Behind the disparate topics on which votes occur is an underlying dimension of international cooperation and conflict that reflects the broader foreign policy positions of states, large and small. We must therefore first construct a measure capable of capturing a latent foreign policy tendency driving a large number of individual choices across a range of topics. Second, our measure must be sensitive to short-term changes in foreign policy that could potentially vary across any changes in leadership. Third, we must establish continuity across periods such that the choices made in the UNGA are comparable from session to session, rather than subject to the varying agenda of that body. For instance, in a given session, there might be many more votes on the Israeli-Palestinian conflict compared to the next or previous session, which could inaccurately be interpreted as a change in individual countries voting patterns. Finally, our measure should detect change in each country without relying on relative similarity to any one country. By tracking a countrys behavior only relative to the U.S., for instance, we lose the ability to account for changes in the U.S. itself. 14
14

Most scholars investigating change in UN voting have employed the affinity score (Gartzke

1998) to assess how alignment with the U.S. changes over time. A problem with this measure, however, is that it does not take into account movement in the U.S. position itself. Like Voeten (2004), we find that the U.S.s UNGA voting behavior has changed over time. For instance, 17

We have created a new measure of revealed foreign policy preferences that attempts to address each of these issues. We begin by estimating the ideal points of states using a spatial voting model, specifically the item response theory model used by Clinton et al. (2004). The spatial voting model is designed to capture underlying tendencies in voting patterns that occur on an overarching dimension of conflict in a choice environment. This method is widely used for uncovering latent ideological tendencies behind voting in legislative or judicial contexts and has previously been applied to the UNGA (e.g. Voeten 2000, 2004, Reed et al. 2008). We employ the Clinton et al. model in conjunction with an approach similar to that of Bailey (2007), who uses votes that occur in multiple sessions to anchor the preference space. This is designed to ensure that choices in different periods can be compared, despite changes in the range of choices from session to session. Bailey uses the information from those questions on which legislators had taken a position in multiple time periods as bridge votes. In a similar vein, we fix the location of those UNGA resolutions that are considered in multiple sessions by considering them as if they were a single item in the voting data. This provides the basis upon which we link together each country-session across time. Except that these votes are fixed i.e., grouped across time into single votes our approach is otherwise a joint scaling of the entire UNGA voting record within our sample, as if each country-session existed simultaneously in a common preference space. This approach is possible because of the high frequency with which the UNGA considers the same resolutions. In the period under study, votes occurred on 293 resolutions that were not George W. Bushs record was quite different from Clintons, and thus changes in similarity between the voting records of other states and the U.S. might not reflect changes in the votes of other states at all. 18

unique. These resolutions were linked across time using the titles obtained from Voeten and Merdzanovic (2008) as follows. First, we aligned each item with the same title. For instance, each time the UNGA voted on a resolution entitled Prevention of an arms race in outer space, we allowed this to function as effectively the same vote such that it would have a fixed location in the preference space. Note that our sample is limited to 1985-2008 because only these resolutions employed a system of UNGA resolution titles that makes this grouping technique possible. To ensure consistency was maintained in the response, we then removed from each set any resolutions for which substantial shifts in the numbers of yeas, nays, or abstentions occurred relative to the adjacent years in that resolution group.15 This process produced 1,137 unique resolutions from a total of 2,099 resolutions. With this unified matrix of votes in place, we estimate the ideal point for each country in each session using a Bayesian MCMC simulation approach (Clinton et al. 2004, Jackman 2004). We focus on the dichotomous choice between supporting resolutions or not supporting them, which can manifest as either abstentions or formal no votes (with absences treated as missing data).16 The result of this process is an estimate of the ideal point for each country in each
15

To be as conservative as possible in treating multiple resolutions as single votes, we removed

dissimilar vote outcomes and allowed multiple groups with the same title when vote distributions were dissimilar. We excluded the resolution from the group if the total of yes/no/abstain votes was greater (less) than the median number of yes/no/abstain plus (minus) the standard deviation of that set within the group. When the group contained more than five resolutions, this was done using the median from the first five resolutions and again from the last five resolutions.
16

As Voeten (2000, 193) explains, "since UNGA resolutions are not binding, what really matters

is whether or not a state is willing to go on the record for supporting a resolution." 19

session that is comparable to their estimates in other sessions (i.e., on a common scale) yet sensitive to any changes in each states pattern of voting relative to other states on the scale.17 From these ideal points, we generate a continuous measure of foreign policy change across leadership periods. We first aggregate the data by averaging the ideal points across sessions for each leadership period. We then construct a measure of UN voting change that represents the absolute value of the difference between the mean ideal point of each leader period and that of the previous leader period. This is our dependent variable in the analysis below. Measuring Changes in Domestic Sources of Leader Support and Regime Type In order to assess the effect of changes in domestic interests on UNGA voting, we need to identify instances in which a leader comes to power who has a different source of leader support (SOLS) than her predecessor.18 The coding rules used to determine whether leadership changes are accompanied by SOLS changes vary depending on regime type.19 Country-years are coded as

17

The anchoring effect of the constraints we introduce provides sufficient consistency to produce

patterns similar to those in earlier work, such as Voeten (2004).

18

To identify these cases, we build on the coding rules developed by Leeds et al. (2009), who

collected data on SOLS changes for countries with bilateral alliances between 1919-2003. We extend their data by including all countries with a population larger than 500,000. Teams of coders at two different universities coded each case. Inter-coder reliability was very high, with agreement in 93.4% of state-years with leadership transitions during 1945-2008.
19

Note that our coding is designed only to measure when a leader is replaced with a leader from

a distinct SOLS, as needed to test the hypotheses above. Given the wide range of domestic 20

democratic if they have a POLITY IV democracy score of 6 or higher and non-democratic otherwise (Marshall, Jaggers, and Gurr 2010).20 In democracies, we conceptualize a leaders SOLS as those who vote for or associate with the leaders party. We follow the basic rule that whenever a leader with a different party affiliation comes to power, this constitutes a SOLS change. In cases where a leader is an independent we follow two rules: 1) if the new leader is the pre-designated successor of the old leader, we code no SOLS change. Pre-designated successors are vice presidents, those appointed by the outgoing leader, or relatives (brother, son) of the old leader; 2) if the voters who elected both leaders are similar based on their regional, ethnic, class etc. characteristics, we also code no SOLS change. In non-democracies, the identification of SOLS changes is more challenging. Here we rely primarily on Geddes, Wright, and Frantzs (2012) classification of post-1945 autocracies as single-party, military, personalist, monarchical systems or combinations thereof. The use of this categorization scheme is appropriate because their coding criteria emphasize control over access to power and influence rather than formal institutional characteristics (Geddes 1999: 123). Geddes et al.s classifications are thus based on identification of the groups that form the

political environments represented in our broad sample of states, it is not possible to capture the specific foreign policy preferences of each SOLS and thus to predict a direction of change.
20

Interruptions (-66) or interregna (-77) are considered to be non-democratic years. Transition

periods (-88) are coded as democratic or non-democratic based on Cheibub et al. (2010). 21

leaders support base and hold sway over policymaking. When these support bases change, we should be more likely to observe shifts in foreign policy.21 Because in single-party systems consecutive leaders are accountable to the party cadre, and because party elites are likely to share basic policy preferences over time, we code no SOLS changes for the duration of a single-party regime. Similarly, we code no SOLS changes for the duration of a single military regime. Studies suggest that military officers share an overarching commitment to advancing the military interest, and military leaders depend heavily on the support of other high-ranking military officers within their regime (Geddes 1999: 126). Thus, different leaders within the same junta should not be expected to pursue different policies. Sometimes, however, one military regime is replaced by a different military regime with a support base that is regionally, ethnically or ideologically distinct. In these instances, we do code a SOLS change. In personalist systems, because a leaders support group consists of a clique of family and friends that are loyal to that particular leader, changes in leadership are usually accompanied by changes in the source of leader support. Exceptions are pre-designated successors who build on the outgoing leaders supporters and should continue to pursue policies that are in the cliques interest. Thus, in personalist systems, we code SOLS changes if a new leader who is not a pre-designated successor assumes office. In monarchies, if a ruler descends from the same dynasty as her predecessor, the source of support is likely to stay the same. We only code SOLS changes in monarchies when a leader transition is associated with the ascendance of a new dynasty. Geddes et al. code some countries as hybrids of two or more of
21

It is worth noting that some of these changes may entail institutional change as well within the

broader category of autocracies. Thus, it is probably not possible to disentangle all forms of institutional changes in autocracies from simultaneous changes in the source of leader support. 22

the above regime types. With some rare exceptions, we code no SOLS changes for the duration of a single hybrid-regime.22 Some country-years cannot be classified according to the Geddes et al. scheme, however. Most of these are country-years during which politics are in flux. Some of these cases are characterized by warlordism (where no particular leader can be said to be in control of the entire country and its policies) or foreign occupation (where no real domestic competition exists because policies are directed by outsiders). We treat cases in which no domestic actor has clear control over the foreign policy apparatus as experiencing no changes in domestic sources of support during the period of incapacity. For countries that have functioning governments but still display too much ambiguity to assign them to any of the Geddes et al. categories, we use a pre-designated successor rule. SOLS changes are coded as occurring when the new leader is not the pre-designated successor of the old leader. Using these coding rules, we create a dummy variable, SOLS change, that is coded 1 if the leaders SOLS is different than her predecessors. Democracy is coded 1 if the country is democratic throughout the entire duration of the current leaders tenure and the entire duration of her predecessors term. Only if the two leaders are operating in a stable democratic regime should we see an attenuated effect of SOLS changes. If the current leader presides over a democracy but her predecessor did not, we would expect a much more significant effect of SOLS change on UN voting.
22

Geddes et al. do not code countries with populations below one million. We categorize nine

countries with populations between 500,000 and one million according to Geddes et al.s coding scheme and code them based on the rules outlined above: Bahrain, Bhutan, Comoros, Cyprus, Djibouti, Equatorial Guinea, Guyana, Qatar, and Solomon Islands. 23

Of our 541 leadership periods, 305 were democratic both during the leaders tenure and her predecessor's, 142 were non-democratic throughout the leaders tenure and her predecessors, and in 94 cases either the current leader or her predecessor governed during a transition from autocracy to democracy or vice versa. We find a greater number of leadership transitions in democracies despite the fact that stable democracies (countries that are democratic throughout our observation period) make up only 31% of countries in our data. Stable democracies have on average 4.6 leaders, while other countries average 3.3 leaders, a statistically significant difference.23 Democracies also experience significantly more SOLS changes than nondemocracies. Among the 305 democratic leadership periods, 202 represent SOLS changes. Among the 236 non-democratic or regime transition leadership periods, 138 represent SOLS changes. Control Variables While our main focus is on the effect of SOLS changes in democracies and nondemocracies, we also conduct analyses including three control variables. Two of these variables capture different aspects of how the Cold War and its ending may have affected UNGA voting. Under the rigid bipolar structure of the Cold War, international pressures to vote consistently with ones bloc constrained a countrys ability to change its positions. In some countries, Cold War politics even constrained domestic change itself. To account for systematic differences
23

29% of countries in our sample were consistently non-democratic, while 41% experienced

regime transitions. The twenty-four year period under observation is an era of significant domestic change as a result of the Third Wave of Democratization as well as some reversals to non-democracy. The average number of leaders in consistently non-democratic countries is 2.2, while it is 4 in countries that experienced regime transitions. 24

between the Cold War and post-Cold War periods, we create a Cold War Period variable that is coded 1 for leaders that end their term before 1990. Only 23 leaders completed their term before the Cold War ended. It is important not only to capture the different structure of international constraints before and after 1990, but also to account for the short-term effect of the end of the Cold War. The fall of the Iron Curtain was a watershed event that led both to significant international realignments and to domestic political changes in a number of countries. To ensure that we do not falsely attribute changes in UNGA voting to SOLS changes, when really both were due to the massive changes in the international system, we include a control for whether the current leader is the first post-Cold War leader. This dummy variable, First Post-Cold War, is coded 1 if the predecessors term was completed before 1990, but the new leaders term included 1990. In the data, 69 leaders are coded as first post-Cold War leaders based on our operationalization. In addition to accounting for the dynamics associated with the Cold War and its termination, we also control for regime change. Regime transitions from democracy to autocracy and vice versa constitute significant changes in a countrys institutional rules that are likely to affect UN voting. Regime transitions also often involve changes in the source of leader support. Thus, in cases of regime change, it is difficult to disentangle the effects of new institutions from the effects of changes in the leaders societal support base. Yet, while regime transitions are usually accompanied by SOLS changes, SOLS changes also occur quite frequently in the absence of this kind of dramatic institutional change.24 By controlling for transitions between
24

We code a SOLS change in 91% of the cases where a leader came to power heading a new

regime type, but only 16% of SOLS changes occurred when there was also a regime change. The correlation between SOLS change and regime change is 0.21. 25

democracy and autocracy, we provide a more conservative test that is better at isolating the pure effects of changes in sources of leader support. Regime change is coded 1 if there is a change in a countrys POLITY IV democracy score from 6 or above to below 6 or vice versa when the leader assumes office. Of the total 541 leader periods, 58 experience such a regime transition. VI. EMPIRICAL RESULTS We begin our analysis with a simple regression model that includes our main theoretical variables of interest, SOLS change and democracy, and the control variables. Our dependent variable, the absolute change in UNGA voting between the current leader and her predecessor, is logged in order to approximate a normal distribution.25 All models also include state-level random intercepts. This allows us to account for unmeasured factors associated with states, while allowing even states with very few leadership changes to have a meaningful basis of comparison. TABLE 1 ABOUT HERE Table 1 shows that, in line with Hypothesis 1, leadership changes that coincide with SOLS changes lead to greater shifts in UN voting than leadership changes in which the new leader has the same societal base of support as her predecessor.26 We also find that, consistent with Hypothesis 2, our democracy variable has a statistically significant negative effect on change in a states UNGA voting. That is, leadership changes in continuing democracies are

25 26

Because some values are near zero, we add a constant (0.01) to this variable before logging. As a robustness check, we used a variable that includes cases where, in parliamentary

democracies, junior partners enter or leave the coalition. Results are similar to those presented. 26

generally less likely to lead to changes in UN alignment than leadership transitions in nondemocracies.27 In order to establish how the SOLS change effect differs between the two regime types, we split the sample between democracies (Model 2) and non-democracies (Model 3). Unlike the democratic sample, which includes only cases in which the current and previous leader both presided over a democracy, the non-democratic sample may include leader transitions that were accompanied by regime changes so we control for regime transitions in this set.28 We find that the effect of SOLS changes is not statistically significant within the sample of democracies. However, SOLS changes have a statistically significant positive effect on change in UNGA voting when the leader transition involves leaders of non-democracies. That is, the overall effect of SOLS changes, shown in Model 1, is being driven mostly by SOLS changes occurring in nondemocracies. Regime transitions that are not SOLS-changes, meanwhile, have no effect in this sample (or in the other models). Hypothesis 3 also implies that democracies should display greater foreign policy stability than non-democracies when leadership changes coincide with changes in the source of leader
27

Recall that democracy in this context refers to cases that were democratic across the period

of change, while non-democratic can refer either to cases where both leaders ruled autocracies or where either leader presided over a regime transition from democracy to autocracy or vice versa.
28

We also conducted analyses where we focused on transitions from one non-democratic leader

to another, excluding leader changes that coincided with regime transitions. The results hold. The results for models 1 & 6 are also very similar whether cases of regime transitions are included or excluded. We are thus confident that we are capturing differences between democracies and nondemocracies rather than between democracies and transitional regimes.

27

support. Models 4 and 5 examine this proposition by splitting the sample into those leadership periods that began with SOLS changes (Model 4) and other leader changes (Model 5). We find that, compared to non-democracies, continuing democracies are significantly less likely to experience shifts in their UN voting patterns when a new leader with a different societal support base comes to power. They are, however, no more stable than non-democracies when leaders with the same domestic support base as their predecessors assume office. While this is consistent with the logic of Hypothesis 3, this suggests that the relative stability of democracies predicted by Hypothesis 2 (and found in Model 1) is driven by those cases in which a change in leadership results in a change in the source of leader support.29 In the final model, Model 6, we examine the size of the difference in the SOLS change effect between regime types, using the full sample and interacting the SOLS change and democracy variables. The interaction term is negative as expected but not statistically significant at conventional levels, indicating that the effect of SOLS change in democracies and nondemocracies is not statistically different. While we cannot reject the hypothesis that SOLS changes have similar mean effects in democracies and non-democracies, it does appear that the effect of SOLS changes is more robust in non-democracies than in democracies. Mirroring the split sample results, the results in Model 6 suggest that we can reject the null hypothesis of no effect of SOLS changes on changes in UNGA voting more confidently in non-democracies (p=0.016) than in democracies (p=0.084). Furthermore, we observe that continuing democracies

29

We also compared variation in UNGA voting within leader terms across regime type. Non-

democratic leaders display significantly more volatility in ideal points over their tenure than democratic leaders, even controlling for the number of years in office.

28

display more stability in the face of SOLS changes (p=0.020) while there is no real difference between democracies and non-democracies when there is no SOLS change. As for the effects of the other control variables described above Cold War Period, and First Post-Cold War, the only one of these variables to have a statistically significant effect on UN voting is First Post Cold War. 30 Models 2 and 3 indicate that this effect is driven by leaders in non-democracies, where leaders that were in power in the immediate aftermath of the Cold War pursued a foreign policy that was distinct from their Cold War predecessors. These leaders, some of whom weathered the transition into the new era and some of whom came to power right after the end of the Cold War, were clearly facing a radically different international environment. As traditional IR theories would predict, the profound changes in the structure of the international system exerted significant influence on a number of countries foreign policy positions. Interestingly, the magnitude of the effect of the end of the cold war on UNGA voting is quite similar to the effect of changes in the domestic sources of leader support. This lends
30

In order to ensure that we have captured most of the exogenous factors behind change in our

dependent variable, we examined a model controlling for the average change in the dependent variable for the year of a leadership change, as well as dummies for each region. Neither of these substantially affect the results presented. We also ran a model controlling for changes in a countrys power. For this measure, we averaged the predecessors composite indicator of national capability (CINC) score over his tenure and then subtracted that quantity from the leaders average CINC score (Singer et al. 1972). There are few cases that experience significant changes in power and this measure is not a significant predictor of UN voting change. Our main conclusions hold when this measure is included, but missing data for the year 2008 leads us to lose observations. 29

support to the notion that changes in the leaders source of societal backing can be important factors in explaining foreign policy change. The overall pattern we uncover is that, in non-democracies, SOLS changes exert a significant effect on UN voting change and non-democratic institutions do little to constrain this effect. In democracies, SOLS changes may also lead to foreign policy change, but this effect appears to be somewhat less significant. Democratic institutions successfully mitigate the effect of SOLS changes on foreign policy and thus promote greater consistency. Our findings are nicely illustrated by two examples from Iran and Australia. One of the more significant shifts in UN voting in our data is connected to the 2005 transition from Khatami to Ahmadinejad in Iran. While Supreme Leader Khamenei possesses ultimate authority over Irans foreign policy, these two presidents were able to leave their own marks on Irans stance in the international arena. The Khatami government generally took a more conciliatory approach to the West and embraced aspects of globalization, while Ahmadinejad is a staunch anti-Americanist, famous for his tirades in the UNGA. As a reformist, Khatami relied on the support of technocrats, younger entrepreneurs, and bazaar (bazaari) merchants and traders pressing for decreased state control, [seeking] privatization, increased trade with Europe and Asia, a utilitarian over an ideological approach to foreign policy, and decreased tensions with international institutions (Solingen 2007, 178) as well as women and the youth who did not share the same revolutionary fervor as their parents. Ahmadinejad, on the other hand, has the support of the Revolutionary Guards, paramilitary youth groups, and social conservatives among the clergy and bazaari. In favor of these groups, Ahmadinejad opted for inward-looking economic policies. Even Irans nuclear policy can be seen as advancing his

30

supporters interests. Sanctions shelter their businesses from competition and allow them to benefit from selling banned products (Solingen 2007, 180-181). By contrast, consider the example of constrained change in a democratic regime seen in the recent politics of Australia. After the 2007 transition from the center-right Liberal Party (John Howard) to the center-left Labor Party (Kevin Rudd), the new government made a noteworthy foreign policy shift by reversing the policy of the previous government and voting against the Israeli position on several UNGA resolutions related to settlements in Palestinian territories (Hudson, 2008). However, Australias overall pattern of voting remained fully consistent with the position of the previous government, including in support of Israel on many other resolutions. VII. CONCLUSION International relations scholarship has traditionally been more successful at explaining foreign policy continuity rather than change. Part of the problem lies in the excessive focus on the structure of the international system and basic state characteristics like power and geographic position. Given the relatively immutable nature of these factors, they explain continuity well but fall short if the goal is to predict change. The basic premise of this paper is that if we want to understand foreign policy change we need to consider domestic political processes. Domestic changes that bring to power new leaders with different societal support groups can be a powerful source of change. If leaders represent different interests and preferences than their predecessors, they may opt to pursue a distinct foreign policy course. Yet, the extent to which such changes in leader support entail shifts in foreign policy should also depend on the domestic institutional context, with democratic institutions mitigating against excessive change and non-democratic institutions allowing new leaders greater leeway in the re-direction of foreign policy.

31

Our empirical focus in this study lies in explaining changes in countries UN voting patterns. Voting in the UNGA is not only a frequently used measure of foreign policy preferences, but it is also particularly useful for gauging the effect of domestic political changes on foreign policy. The opportunity to vote in the UNGA occurs under virtually every domestic leader, and voting is a monadic activity. Leaders are constrained by international law when it comes to changing policies enshrined in treaties, but there are no such constraints on UN voting. This suggests that new leaders have more leeway in changing foreign policy positions in the UN than elsewhere. It also means that any stability that we observe is likely to have been induced by domestic institutions rather than international ones. Using a new measure of change in UNGA voting over the 1985-2008 time period and new data on changes in leaders sources of domestic support, we find that SOLS changes are consistently a significant predictor of foreign policy change in non-democracies but not in democracies. Democracies also display greater foreign policy consistency in the face of SOLS changes than non-democracies. Interestingly, however, it appears that the difference between democracies and non-democracies is not as pronounced in this relatively unconstrained area of foreign policy as a previous study has shown it to be in the highly legalized arena of military alliances (Leeds, Mattes, and Vogel 2009). Democracies seem to be more responsive to shifting domestic preferences in areas of foreign policy that are subject to fewer domestic and international political constraints, but, even in these areas, democracies do display a greater foreign policy consistency than non-democracies. Leaders in democratic regimes may therefore not face a dilemma between responsiveness to domestic interests and maintaining stable foreign policy.

32

Our study supports a perspective that views foreign policymaking even in the high politics arena as intertwined with the domestic policymaking process, whereby groups compete to advance their interests and implement their preferred policies within given institutional rules. The fact that foreign policy generally involves strategic interaction with other states does affect both domestic competition over foreign policy and ultimate policy choices, but politics does not entirely stop at the waters edge. This suggests that policymakers hoping to predict foreign policy change should pay particular attention to the dynamics of domestic political competition and leadership selection, particularly in non-democracies.

33

REFERENCES Alker, Hayward and Bruce Russett. 1965. World Politics in the General Assembly. New Haven, CO.: Yale University Press. Bailey, Michael A. 2007. Comparable Preference Estimates across Time and Institutions for the Court, Congress, and Presidency. American Journal of Political Science 51(3): 433-448. Bueno de Mesquita, Bruce, Alastair Smith, Randolph M. Siverson, and James D. Morrow. 2003. The Logic of Political Survival. Cambridge, MA: The MIT Press. Carter, David B. and Randall W. Stone. 2011. Democracy and Multilateralism: The Case of Vote Buying in the UN General Assembly. Unpublished Manuscript available at http://www.rochester.edu/college/psc/stone/working_papers/CarterandStone2012.pdf (accessed April 20, 2012). Cheibub, Jose Antonio, Jennifer Gandhi, and James Vreeland. 2010. Democracy and Dictatorship Revisited. Public Choice 144 (1-2): 67-101. Clinton, Joshua, Simon Jackman, and Douglas Rivers. 2004. The Statistical Analysis of Roll Call Data. American Political Science Review 98 (2): 35570. Davis, David R. and Will H. Moore. 1997. Ethnicity Matters: Transnational Ethnic Alliances and Foreign Policy Behavior. International Studies Quarterly 41 (1): 171-184. Dominguez, Jorge. 1978. Cuba: Order and Revolution. Cambridge, MA: Belknap Press of Harvard University Press. Dreher, Axel and Nathan Jensen. 2012. Country or Leader? Political Change and UN General Assembly Voting. Unpublished Manuscript, available at http://www.pubchoicesoc.org/papers_2012/Dreher_Jensen.pdf (accessed April 20, 2012).

34

Dreher, Axel, Peter Nunnenkamp, and Rainer Thiele. 2008. Does U.S. Aid Buy UN General Assembly Votes? A Disaggregated Analysis. Public Choice 136 (1): 139-164. Dreher, Axel and Jan-Egbert Sturm. 2012. Do the IMF and the World Bank Influence Voting in the UN General Assembly? Public Choice 151 (1): 363-397. Gartzke, Erik. 1998. Kant We All Just Get Along?: Opportunity, Willingness, and the Origins of the Democratic Peace. American Journal of Political Science 42 (1): 1-27. Geddes, Barbara. 1999. What Do We Know About Democratization after 20 Years? Annual Review of Political Science 2: 115-144. Geddes, Barbara, Joseph Wright and Erica Frantz. 2012. Authoritarian Regimes: A New Data Set. Manuscript. Available at http://dictators.la.psu.edu/ (accessed May 10, 2012). Goemans, Henk E., Kristian Skrede Gleditsch, and Giacomo Chiozza. 2009. Introducing Archigos: A Data Set of Political Leaders. Journal of Peace Research 46 (2): 269-283. Hagan, Joe D. 1989. Domestic Political Regime Change and Third World Voting Realignments in the United Nations, 1946-1984. International Organization 43 (3): 505-541. Hudson, Phillip. 2008. UN Vote: Hudson Breaks with Rudd on Israel. Sydney Morning Herald, November 10. Available at http://www.smh.com.au/articles/2008/11/09/1226165386581.html (accessed July 11, 2012). Jackman, Simon. 2004. Bayesian Analysis for Political Research. Annual Review of Political Science 7: 483505. Kim, Soo Yeon and Bruce Russett. 1996. The New Politics of Voting Alignments in the United Nations General Assembly. International Organization 50 (4): 629-52.

35

Lai, Brian and Daniel S. Morey. 2006. Impact of Regime Type on the Influence of U.S. Foreign Aid. Foreign Policy Analysis 2 (4): 385-404. Leeds, Brett Ashley, Michaela Mattes, and Jeremy S. Vogel. 2009. Interests, Institutions, and the Reliability of International Commitments. American Journal of Political Science 53 (2): 461-476. Lobell, Steven E. 2004. Politics and National Security: The Battles for Britain. Conflict Management and Peace Science 21 (4): 269-286. Marshall, Monty G., Keith Jaggers, and Ted Robert Gurr. 2010. Polity IV Project: Dataset Users Manual. College Park, MD: University of Maryland. Available at http://www.systemicpeace.org/inscr/p4manualv2010.pdf (accessed April 20, 2012). Martin, Lisa L. 2000. Democratic Commitments. Princeton, NJ: Princeton University Press. McGillivray, Fiona and Alistair Smith. 2008. Punishing the Prince: A Theory of Interstate Relations, Institutions, and Leader Change. Princeton, N.J.: Princeton University Press. Milner, Helen V. 1999. The Political Economy of International Trade. Annual Review of Political Science 2: 91-114. Moon, Bruce E. 1985. Consensus or Compliance? Foreign Policy Change and External Dependence. International Organization 39 (2): 297-329. Moore, John Allphin and Jerry Pubantz. 2008. Encyclopedia of the United Nations, Volume 1. 2nd edition. New York: Facts on File. Narizny, Kevin. 2003. Both Guns and Butter, or Neither: Class Interests in the Political Economy of Rearmament. American Political Science Review 97 (2): 203-220. _________. 2007. The Political Economy of Grand Strategy. Ithaca, NY: Cornell University Press.

36

Potrafke, Niklas. 2009. Does Government Ideology Influence Political Alignment with the U.S.? An Empirical Analysis of Voting in the UN General Assembly. The Review of International Organizations 4 (3): 245-268. Ratner, Ely. 2009. Reaping What You Sow. Democratic Transitions and Foreign Policy Realignment. Journal of Conflict Resolution 53 (3): 390-418. Reed, William, David H. Clark, Timothy Nordstrom, and Wonjae Hwang. 2008. War, Power and Bargaining. Journal of Politics 70 (4): 1203-1216. Singer, J. David, Stuart Bremer, and John Stuckey. 1972. Capability Distribution, Uncertainty, and Major Power War, 1820-1965. In Bruce Russett (ed) Peace, War, and Numbers, Beverly Hills, CA: Sage, 19-48. Solingen, Etel. 2007. Nuclear Logics: Contrasting Paths in East Asia and the Middle East. Princeton, NJ: Princeton University Press. Tsebelis, George. 2002. Veto Players: How Political Institutions Work. Princeton, NJ: Princeton University Press. Vengroff, Richard. 1976. Instability and Foreign Policy Behavior: Black Africa in the UN. American Journal of Political Science 20 (3): 425-438. Voeten, Erik. 2000. Clashes in the Assembly. International Organization 54 (2): 185-215. _________. 2001. Domestic Institutional Change and the External Orientation of States. Georgetown University: Unpublished Manuscript. _________. 2004. Resisting the Lonely Superpower: Responses of States in the United Nations to U.S. Dominance. Journal of Politics 66 (3): 729-754.

37

Voeten, Erik and Adis Merdzanovic. 2008. United Nations General Assembly Voting Data. Available at http://hdl.handle.net/1902.1/12379 UNF:3:Hpf6qOkDdzzvXF9m66yLTg== V1 Wang, T.Y. 1999. U.S. Foreign Aid and UN Voting: An Analysis of Important Issues. International Studies Quarterly 43 (1): 199-210.

38

Table 1. Change in UNGA Voting Between Leaders as a Function of SOLS Change and Regime Type, 1985-2008 (1) All Cases (2) Dem. (3) Non-Dem. (4) SOLS Changes (5) Non-SOLS Leader Changes (6) All Cases

SOLS Change Democracy Democracy X SOLS Change Cold War Period First Post-Cold War Regime Transition Constant Observations Number of Countries

0.229*** (0.081) -0.210** (0.089)

0.176 (0.111)

0.274** (0.118) -0.285*** (0.109) 0.185 (0.316) 0.512*** (0.159) 0.187 (0.146) -1.623*** (0.107) -0.009 (0.351) 0.128 (0.185) 0.116 (0.164) -1.252*** (0.093) -0.122 (0.146) 0.047 (0.352) 0.398** (0.165) 0.404 (0.446) -1.619*** (0.120) 201 102

0.016 -0.269 (0.237) (0.319) 0.253* -0.097 (0.134) (0.222) 0.179 (0.148) -1.554*** -1.687*** (0.091) (0.096)

0.282** (0.117) -0.160 (0.133) -0.090 (0.161) 0.016 (0.237) 0.250* (0.135) 0.156 (0.148) -1.579*** (0.103) 541 147

541 305 236 340 147 82 100 115 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

39